Skip to main content
Internet Archive's 25th Anniversary Logo

Full text of "The Dark Side of Galaxy Color"

See other formats

Mon. Not. R. Astron. Soc. 000, 000-000 (0000) Printed 23 April 2013 (MN WTeK style file v2.2) 

The Dark Side of Galaxy Color 

Andrew P. Hearin^ and Douglas F. Watson^, ^ 

^ Fermilab Center for Particle Astrophysics, Fermi National Accelerator Laboratory, Batavia, IL, 60510-0500 

'^NSF Astronomy & Astrophysics Postdoctoral Fellow, Department of Astronomy & Astrophysics, The University of Chicago, Chicago, IL 60637 

^ Kavli Institute for Cosmological Physics, 564-0 South Ellis Avenue, The University of Chicago, Chicago, IL 60637 














23 April 2013 


We present age distribution matching^ a new theoretical formalism for pre- 
dicting how galaxies of luminosity L and color C occupy dark matter halos. Our 
model supposes that there are just two fundamental properties of a halo that 
determine the color and brightness of the galaxy it hosts: the maximum circular 
velocity Vmax, and the redshift at which the galaxy will be starved of cold gas 
^starve- The halo property ^starve is intended to encompass physical characteris- 
tics of a halo's mass assembly history (MAH) that may deprive the galaxy of 
its cold gas supply and, ultimately, quench its star formation. These include: 
(a) -Zacc, the epoch a halo accretes onto a larger halo, (b) Zchar, the epoch a 
halo reaches a characteristic mass (~ lO^^/i~^M0), and (c) Zform, the epoch a 
halo transitioned from the fast- to slow-accretion regime. A halo's ^starve value 
is determined by whichever of these events happens first in its MAH. The new, 
defining feature of the model is that, at fixed luminosity, galaxy color is in mono- 
tonic correspondence with Zgtarvc, with the larger values of Zgtarve being assigned 
redder colors. We populate the Bolshoi N- body simulation with a mock galaxy 
catalog based on age distribution matching, and show that the resulting mock 
galaxy distribution accurately describes the luminosity- and color-binned two- 
point correlation function of galaxies in the Sloan Digital Sky Survey (SDSS), 
as well as a variety of low-redshift galaxy group statistics. Our study of the 
property Zstarvo has important implications for how halo MAH influences stellar 
mass assembly. For example, the significance of the epoch Zform suggests a new, 
independent channel for so-called "mass quenching" based on the rate of halo 
mass accretion. We make publicly available our low-redshift, SDSS M^ < — 19 
mock galaxy catalog, and main progenitor histories of all z = Bolshoi halos, 
at littp : / /logrus . uchicago . edu/ '^aphearin/ 

Key words: cosmology: theory — dark matter — galaxies: halos — galaxies: 
evolution — galaxies: clustering — large-scale structure of universe 


Galaxies are not color blind in how they congregate with 
one another. Redder galaxies (older with little active star 
formation) tend to occupy more dense environments, and 
conversely, bluer galaxies (younger with active ongoing 
star formation) preferentially reside in underdense re- 
gions (Balogh et al. 1999; Blanton et al. 2005; Wein- 
mann et al. 2006, 2009; Peng et al. 2010, 2012; Wetzel 
et al. 2011; Carollo et al. 2012). The color dependence of 
galaxy location within the cosmic web also manifests in 
measurements of two-point statistics (e.g., Norberg et al. 
2002; Zehavi et al. 2002, 2005; Li et al. 2006; Yang et al. 
2011; Zehavi et al. 2011; Mostek et al. 2012). Addition- 

ally, there is a clear bimodality in the distribution of 
galaxy color and brightness, with distinct red and blue 
sequences in color-magnitude space, along with a minor- 
ity 'green valley' population (Blanton et al. 2003; Baldry 
et al. 2004; Blanton et al. 2005). This segregation between 
blue star-forming and red passively evolving galaxies has 
been shown to persist out to z ~ 1 (Bell et al. 2004; 
Cooper et al. 2006, 2012) and even as high as z ~ 3 
(Whitaker et al. 2011). While these correlated trends in 
color-magnitude and configuration space are observation- 
ally well-established, our theoretical understanding of the 
physical processes driving these correlations remains in- 

In the standard ACDM cosmological model struc- 

© 0000 RAS 

A. P Hearin & D. F Watson 

tures assemble hierarchically, wherein high density re- 
gions condense and virialize, forming bound structures 
known as dark matter halos. Halos of sufficient mass are 
the natural sites for luminous galaxies to form (White & 
Rees 1978; Blumenthal et al. 1984), and the way galax- 
ies connect to dark matter halos provides the fundamen- 
tal link between predictions of galaxy formation theory, 
such as color, and the concordance ACDM model. Under- 
standing this link requires a detailed knowledge of the 
spatial distribution of galaxies and halos, and towards 
this end large galaxy redshift surveys such as the Sloan 
Digital Sky Survey (SDSS: York et al. 2000) have pro- 
duced detailed three-dimensional maps of galaxies in the 
local Universe. 

Extensive effort has been put forth to understand 
how, as a function of color, galaxies map to dark matter 
halos in a cosmological context, and there are two general 
approaches to constructing the galaxy-halo connection. 
First, there is the halo occupation distribution (HOD) 
technique (and the closely related conditional luminosity 
function technique (CLF: e.g., Yang et al. 2003; van den 
Bosch et al. 2007)), which adopts parametric forms for 
the biased relation between galaxies and halo mass (e.g.. 
Peacock & Smith 2000; Seljak 2000; Scoccimarro et al. 
2001; Berlind & Weinberg 2002; Cooray & Sheth 2002; 
Zheng et al. 2007; Leauthaud et al. 2011). The HOD de- 
scribes this relation by specifying the probability distri- 
bution that a halo of mass M contains A'^ galaxies, along 
with a prescription for the spatial distribution of galax- 
ies within halos. In this formalism, it is typically assumed 
that a single 'central' galaxy lives at the center of each 
distinct halo, with additional 'satellite' galaxies tracing 
the dominant dark matter component. The most com- 
mon implementations of this approach do not explicitly 
use any information about dark matter subhalos (self- 
bound, dark matter structures orbiting within their host 
halo potential). 

The second technique is known as abundance match- 
ing (Kravtsov et al. 2004; Vale & Ostriker 2004; Tasit- 
siomi et al. 2004; Vale & Ostriker 2006; Conroy et al. 
2006; Conroy & Wechsler 2009; Guo et al. 2010; Simha 
et al. 2010; Neistein et al. 2010; Watson et al. 2012; Red- 
dick et al. 2012; Rodriguez-Puebla et al. 2012; Hearin 
et al. 2012; Kravtsov 2013). This is a simple, yet remark- 
ably powerful approach wherein by assuming a mono- 
tonic relation between maximum circular velocity Vmax 
(or halo mass) and luminosity L, (or stellar mass) one 
can 'abundance match' to make the correspondence be- 
tween a dark matter (sub)halo and the galaxy residing 
at its center. In this case, galaxies are assigned to dark 
matter halos with the most massive galaxies residing in 
the most massive halos. This yields an implicit relation- 
ship between L and Vmax such that, by construction, the 
luminosity function of mock galaxy catalogs is in exact 
agreement with the data. 

The most accurate and precise theoretical descrip- 
tions of the galaxy color-halo connection to date have 
been predominantly derived from modeling the color- 
dependent galaxy two-point correlation function (2PCF) 
through the HOD formalism (e.g., Zehavi et al. 2005, 
2011; Skibba & Sheth 2009; Tinker & Wetzel 2010; 
Krause et al. 2013), or semi-analytic models (White & 

Frenk 1991; KaufFmann et al. 1999). However, abun- 
dance matching type theories implemented in a cos- 
mological N-body simulation are now being explored 
(Gerke et al. 2012; Masaki et al. 2013). Alternatively, new 
phenomenological-based approaches that seek to simplify 
the complexities of semi-analytical models are actively 
being pursued (Peng et al. 2010; Mutch et al. 2013; Lilly 
et al. 2013). 

Our intent in this paper is to try and uncover the 
simplest, physically-motivated model for associating the 
color of a galaxy with the mass accretion history of 
the galaxy's parent halo. To that end, we introduce a 
new theoretical formalism which we call age distribution 
matching. This model supposes that there are just two 
fundamental properties of a halo that determine the color 
and brightness of the galaxy it hosts, (1) the maximum 
circular velocity of a test particle orbiting in its gravita- 
tional potential well (i.e., Vmax,), and (2) the redshift at 
which the galaxy will be starved of cold gas, which we 
refer to as ^starve throughout the paper. This halo prop- 
erty Zstarvo is intended to encompass characteristics of a 
halo's assembly history which should be associated with 
physical processes that can deprive a galaxy of a cold 
gas supply and, ultimately, quench star formation. The 
novel feature of the model is that, at fixed luminosity, 
galaxy color is in monotonic correspondence with Zstarvc. 
By constructing a mock galaxy catalog with both r-band 
magnitudes (luminosities) Mr and (g — r) colors, we com- 
pare the predictions of our model to a number of observed 
galaxy statistics. 

The rest of this paper is organized as follows. In § 2 
we discuss the data incorporated throughout this work 
and in § 3 we discuss the simulation, halo catalogs, and 
merger trees used in this study. In § 4 we describe our 
algorithm for constructing our SDSS-based mock catalog 
including a detailed description of our age distribution 
matching formalism. Results are presented in § 5 followed 
by a discussion in § 6. Finally, in § 7 we give a summary 
of our work. Throughout we assume a flat ACDM cosmo- 
logical model with fim = 0.27 and all quoted magnitudes 
use ft = 1. 


We construct our low-redshift mock galaxy sample based 
on the galaxy catalog described in § 2.1. In particular, 
this is the sample that deflnes the distribution of colors 
at flxed luminosity, Psdss(5 — A^)^ required in age dis- 
tribution matching (see § 4.1.3). In § 2.2 we describe the 
projected 2PCF measurements that we use to compare 
with the clustering predictions of our mock sample. We 
use the galaxy group sample described in § 2.3 to test 
our predictions for how galaxies arrange themselves into 
groups by color. 

2.1 SDSS Galaxy Sample 

For our baseline galaxy sample we use a volume-limited 
catalog of galaxies constructed from the Main Galaxy 
Sample of Data Release 7 (Abazajian et al. (2009), DR7 
hereafter) of SDSS. This galaxy catalog is an update 

© 0000 RAS, MNRAS 000, 000-000 

The Dark Side of Galaxy Color 3 

of the Berlind et al. (2006) sample, which was based 
on SDSS Data Release 3, to which we refer the reader 
for details. Briefly, our volume-limited spectroscopic sub- 
sample (T4ff = 5.8 X l(fh~^Mpc^) spans the redshift 
range 0.02 < z < 0.068 with r-band absolute magnitude 
Mr — 5 log ft < —19. Throughout this paper we refer to 
this as the "Mrl9" galaxy sample. 

The term "fiber collisions" refers to cases when the 
angular separation between two or more galaxies is closer 
than the minimum separation permitted by the plugging 
mechanism of the optical fibers in the spectrometer (see 
Guo et al. (2012), and references therein). The treatment 
of fiber collisions in the DR7-based galaxy sample we use 
in this paper difl'ers from that in the catalog presented in 
Berlind et al. (2006). These differences were discussed in 
detail in Appendix B of Hearin et al. (2012), to which we 
refer the reader for details. 

2.2 Clustering Measurements 

We compare our model predictions for the color- 
dependent 2PCF to the SDSS measurements presented 
in Zehavi et al. (2011). Using the observed (g — r) colors, 
they separate red and blue galaxy populations based on 
the following magnitude-dependent color cut: 

0.21 - 0.03M,. 


We use the luminosity-binned clustering data presented 
in their Tables C9 & CIO to test our model 2PCF pre- 
dictions for red and blue galaxies. 

2.3 Galaxy Group Measurements 

Galaxy groups are constructed from the Mrl9 galaxy 
sample described in § 2.1 according to the methods dis- 
cussed in Berlind et al. (2006). Briefly, groups are identi- 
fied via a redshift-space friends-of-friends algorithm that 
takes no account of member galaxy properties beyond 
their redshifts and positions on the sky. For each group, 
we define the central galaxy to be the brightest non-fiber- 
collided member, and the group's satellites to be the re- 
maining non-fiber-collided members. The details of how 
fiber collisions influence central/satellite designations in 
this group sample was studied extensively in Hearin et al. 
(2012, hereafter H12), and we follow their conventions 
here. We refer the reader to Berlind et al. (2006) for fur- 
ther details on the group flnding algorithm. 


Throughout this work we use the high resolution, coUi- 
sionless, cold dark matter TV— body Bolshoi simulation 
(Klypin et al. 2011). The simulation has a volume of 
250^ h~'^Mpc^ with 2048^ particles and a cold dark mat- 
ter (ACDM) cosmological model with Qm = 0.27, SIa ~ 
0.73, ^b = 0.042, h = 0.7, as = 0.82, and n^ = 0.95. 
Bolshoi has a mass resolution of 1.9 x 10* Mq and force 
resolution of 1/i^^kpc, and particles were tracked from 
z = 80 to z = 0. Bolshoi was run with the Adaptive 
Reflnement Tree Code (ART; Kravtsov et al. 1997; Got- 
tloeber & Klypin 2008). The Bolshoi data are available 

at and we refer the reader 
to Riebe et al. (2011) for additional information. 

Halos and subhalos were identifled with the phase- 
space temporal halo finder ROCKSTAR (Behroozi et al. 
2013a,b), capable of resolving Bolshoi halos and subhalos 
down to Vmax ~ 55km s~^. Halo masses were calculated 
using spherical overdensities according to the redshift- 
dependent virial overdensity formalism of Bryan & Nor- 
man (1998). We use the z = ROCKSTAR halo catalog 
and merger trees to model the galaxy color-halo connec- 


In this section we describe our algorithm for construct- 
ing our mock sample of galaxies and galaxy groups. We 
apply age distribution matching, defined in § 4.1, to the 
ROCKSTAR halo catalogs at z = to construct a sam- 
ple of mock galaxies with both r-band luminosities Mr 
and {g — r) colors. After constructing the mock galaxy 
sample from the halo catalog, we place the mock galax- 
ies into redshift space and apply the same group-finder 
to the mock data that was applied to the SDSS Mrl9 
galaxy sample (see § 2.3). We now discuss the steps of 
this procedure in turn. 

4.1 Age Distribution Matching 

4-1.1 Overview 

Age distribution matching is a two-phase algorithm for 
assigning both luminosity and color to the galaxies lo- 
cated at the center of (sub)halos in an N-body simulation. 
First, luminosities are assigned to (sub)halos using abun- 
dance matching (described in § 1 and below in § 4.1.2). 
Second, the mock galaxies are binned by luminosity, and 
colors are assigned to the mock galaxies by randomly 
drawing from the color PDF Psdss(<7 — 'r\L) defined by 
our SDSS Mrl9 galaxy sample. The random draws from 
^SDSs(3 — r\L) are rank-ordered, and the mock galax- 
ies in each luminosity bin are rank-ordered by the halo 
property ^starve (see section 4.1.3), such that at each lu- 
minosity mock galaxies where starvation occurs earlier 
are assigned redder colors. We refer the reader to § 6 for 
a discussion of the physical motivation and implications 
of the model we introduce. 

4-1.2 Luminosity Assignment 

We assign r-band luminosity to Bolshoi (sub)halos using 
the standard abundance matching technique. This tech- 
nique has been widely used in the literature and so we do 
not discuss all of the technical details of this phase of our 
algorithm. Our implementation of the abundance match- 
ing luminosity assignment is identical to that described 
in detail in Appendix A of H12, and thus we restrict our- 
selves to an overview here. 

The quantity Vmax = Max | ^GM{< r)/r\ , where 
M(< r) is the mass enclosed within a distance r of the 
halo center, defines the maximum circular velocity of a 
test particle orbiting in the halo's gravitational potential 

© 0000 RAS, MNRAS 000, 000-000 

A. P Hearin & D. F Watson 
All Galaxies 


0.01 : 


/..v.- I 

0.00 1 ■■■■' ............. ■■ .-■„ 

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 


0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 





= ^ 









r : : 


L ,.-r" ' \ 






0.00 1 .^^:^^^^^^^ 

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 


0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 


Figure 1. Galaxy color PDFs from our mock catalog as compared to those measured in the Mrl9 SDSS galaxy catalog. By 
construction, the (g — r) probability distribution of our mock galaxies is in exact agreement with the data (black dotted histograms) 
at all luminosities. 

well. The abundance matching algorithm for luminosity 
assignment assumes a monotonic relationship between lu- 
minosity and Vmax, such that the cumulative abundance 
of SDSS galaxies brighter than luminosity L, ng{> L), 
is equal to the cumulative abundance of halos with cir- 
cular velocity larger than Vmax, nh{> Vmax)- Following 
common practice we abundance match on the property 
Vpcak, the maximum value Vmax ever attains over the 
entire merger history of the (sub)halo (see Reddick et al. 
2012, for a thorough discussion of the use of Vpeak in 
abundance matching). 

Recent results (Wetzel & White 2010) indicate that 
halos at the faint end of our mock galaxy sample {Mr ~ 
— 19) may be pushing the resolution limit of the Bolshoi 
simulation. We do not attempt to correct for possible 
resolution effects, but note that such corrections (e.g., 
Moster et al. 2010) may be important for precision pre- 
dictions of the small-scale 2PCF. A systematic investiga- 
tion of these resolution effects is beyond the scope of this 
paper, and so we leave this as a task for future work. 

Galaxy formation is a complex process, and so a sin- 
gle halo property such as Vmax cannot uniquely spec- 
ify the luminosity of the halo's galaxy. For example, 
the TuUy-Fisher relation has intrinsic scatter (Pizagno 
et al. 2007), and so abundance matching cannot repro- 
duce the statistics of the observed galaxy distribution 

without scatter (see Behroozi et al. 2010, for a compre- 
hensive analysis of sources of scatter in abundance match- 
ing modeling) . Our luminosity abundance matching tech- 
nique results in ~ O.lSdex of scatter over the luminosity 
range of our sample. Details of our implementation of 
scatter can be found in Appendix A of H12. 

4-1.3 Color Assignment 

Once luminosities have been assigned to the mock galax- 
ies, we proceed to the second phase of our algorithm. 
First, we divide the mock and SDSS galaxies into ten 
evenly spaced bins of r-band magnitude. We use the ob- 
served colors of the SDSS Mrl9 galaxies in each luminos- 
ity bin to define the probability distribution Psdss(5 — 
r\L). If there are A'l mock galaxies in luminosity bin L, 
we randomly draw A^l times from Psdss(5 — tIL) and 
rank-order the draws, reddest first. We assign these col- 
ors to the mock galaxies in luminosity bin L after first 
rank-ordering these A^l mock galaxies by the property 
^starve (defined below), largest first. 

By repeating the above procedure in each luminosity 
bin, the color distribution of the mock galaxies is in ex- 
act agreement with the data, as can be seen in Figure 1. 
Note that this desirable property of our mock would be 
true regardless of the rank-ordering on Zstarvc- The effect 

© 0000 RAS, MNRAS 000, 000-000 

£. 10.0 



^ 100.0 

£ 10.0 



^ 100.0 

£ 10.0 


S:"^ 1.0 




-20 < M, < -21 

-21 < M, < -22 

rp [h-^ Mpc] 

RED • 


The Dark Side of Galaxy Color 5 


10.0 0.1 

rp [h-^ Mpc] 


Figure 2. Luminosity- and color-dependent clustering as predicted by our age distribution matcliing formalism. Left Column: The 
luminosity-binned projected 2PCF predicted by our model (black solid curves) against the clustering exhibited by SDSS galaxies. 
Right Column: In bins of luminosity, we plot the projected 2PCF of red (blue) mock galaxies with red (blue) solid curves. Red, 
filled (blue, open) points show the clustering of red (blue) SDSS galaxies. We use the results presented in Zehavi et al. (2011) for 
all SDSS measurements shown in this figure. 

© 0000 RAS, MNRAS 000, 000-000 

6 A. P Hearin & D. F Watson 

of the rank-ordering is to introduce, at fixed luminosity, 
a correlation between galaxy color and the epoch of star- 

The property Zstarvc is based on three characteristic 
epochs of a halo's mass assembly history. These include: 

!• -Zchar : The first epoch at which halo mass exceeds 
IO^^/i^^A'/q. For halos that never attain this mass Zchar = 

2. ^acc : For subhalos, Zacc is the epoch after which the 
object always remains a subhalo. For host halos, Zacc = 0. 

3 Zform : Using the methods of Wechsler et al. (2002) 
we identify the redshift at which the halo transitions from 
the fast- to slow-accretion regime, as this is the epoch 
after which dark matter ceases to accrete onto the halo's 
central region. 

From these three characteristic epochs we define the red- 
shift of starvation: 


iViaX -(^2acc, •S^char; ^form J ■ 


We provide a full discussion of the physical motiva- 
tion and interpretation of ^starve in § 6. In Appendix A, 
we discuss the details of how each of the above properties 
are computed from Bolshoi merger trees. 

4.2 Group Identification 

Once colors and luminosities have been assigned to the 
halo catalog, we place our mock galaxies into redshift 
space and run the same group-finding algorithm on the 
mock that was run on the SDSS Mrl9 galaxy sample. In 
this way, our mock groups inherit the same purity and 
incompleteness systematics as our SDSS groups. Finally, 
we introduce fiber collisions into our mock sample accord- 
ing to the prescription adopted in H12; this method pre- 
serves the relationship between local galaxy density and 
the likelihood that a fiber collision occurs in the SDSS 
Mrl9 sample. 


We now investigate how well our model can predict a vari- 
ety of observed galaxy statistics. We first demonstrate the 
accuracy of our base Mrl9 SDSS catalog at reproducing 
the observed SDSS luminosity-binned 2PCF, explicitly 
showing that our mock catalog naturally inherits the suc- 
cesses of abundance matching. We then compare the re- 
sulting color-binned 2PCF predictions to SDSS measure- 
ments. Finally, we test the success of our model against 
a variety of galaxy group statistics measured from the 
Mrl9 group catalog. 

5.1 Luminosity- and Color-Dependent 

We begin by showing that our abundance matching-based 
luminosity assignment (described in § 4.1.2) accurately 
predicts the observed luminosity-binned 2PCF as mea- 
sured by Zehavi et al. (2011). To do this, we take our 
volume-limited Mrl9 catalog and sub-divide into three 

luminosity bins: -19 < Mr < -20, -20 < Mr < -21, 
and —21 < Mr < —22. For each of these luminosity bins, 
we compute the real-space 2PCF. We then convert to 
projected space, ■u;p(rp), via: 

mp(r-p) = 2 


+ TV^)d-K 


and use the same upper limit of integration as used in 
these observational samples. 

The black solid curves in the left column of Figure 2 
show the 2PCF as predicted by our mock catalog. There 
is excellent agreement with the Zehavi et al. (2011) SDSS 
measurements (black data points) at each luminosity bin 
and at all projected separations from the linear to highly 
non-linear regimes. 

From this success, we now turn to the right column 
of Figure 2 showing the color-dependent 2PCFs in dis- 
tinct r-band luminosity bins as predicted by our age dis- 
tribution matching formalism. Red filled and blue open 
circles in all panels represent the red and blue galaxy 
populations from Zehavi et al. (2011), respectively. Red 
and blue solid curves are the model predictions. While 
there is some discrepancy between the clustering of blue 
galaxies in the model and the data at large scales for the 
— 20 < Air < —21 bin, overall there is good agreement 
for each luminosity bin and at all Vp separation scales. 
This result is very encouraging since the color-dependent 
2PCF encodes rich information about the galaxy-halo 
connection (e.g.. More et al. 2013; Cacciato et al. 2013). 
However, as shown in Masaki et al. (2013) the 2PCF alone 
is not enough to discriminate between competing models 
for the relationship between a halo and the color of the 
galaxy it hosts. We address this in the next section with 
our exploration of galaxy group statistics. 

We emphasize that we have done no parameter fit- 
ting to achieve the agreement between the predicted and 
measured color-binned clustering. Our algorithm for color 
assignment has no explicit dependence on halo position; 
rather, the clustering signal in our mock emerges as a 
prediction of our theory. Since the primary aim of this 
paper is to introduce the age distribution matching for- 
malism, we have opted for the simplest implementation of 
this framework rather than attempting to finely tune our 
model according to the measured clustering (see § 6 for 
a discussion of more complex possible implementations) . 
The success of our model's prediction for the color-binned 
2PCF is rather striking given the simplicity of the formu- 
lation we present here. 

5.2 Galaxy Group Statistics 

In this section we test how well our model distinctly as- 
signs colors to satellite and central galaxies in our mock 
catalog compared to those colors in the SDSS galaxy 
group catalog. In the top row of Figure 3, we show the 
mean {g — r) color of group galaxies as a function of the 
brightness of the group's central galaxy, M^'^'^ . From 
left to right, we show the results for central galaxies and 
satellites, respectively, in the mock and the SDSS group 
catalog. The dashed line traces the mean {g — r) color 
of the mock galaxies in a given Mr bin. The bottom 
row is similar to the top, except we plot -FBLUE(i), the 

© 0000 RAS, MNRAS 000, 000-000 

The Dark Side of Galaxy Color 7 










0.0 L 




-20.0 -20.5 





Figure 3. Top Row: Mean {g — r) color as a function of absolute magnitude of the brightest central galaxy, Mf^ for central 
galaxies (left panel) and satellite galaxies (right panel). SDSS data are shown as the black filled circles and the dashed curves 
represent the predictions from our mock catalog. Solid grey bands are the errors computed as the dispersion over 1000 bootstrap 
realizations of the galaxy sample (see § 5.2 for details). Bottom Row: Same as the top row but now with the fraction of galaxies 
in groups of a given AI^^'^ which are blue, FbluEj as defined by the color cut in Eq. 1. 

fraction of galaxies that are designated 'blue' according 
to the color cut defined by Eq. 1. 

In all panels of Fig. 3, the solid grey region shows the 
errors on the mean. We use standard bootstrap resam- 
pling techniques to estimate the errors in all of our group- 
based statistics. We compute the errors as the dispersion 
over 1000 bootstrap realizations of the galaxy sample, 
where each realization is constructed by randomly select- 
ing^ Ng galaxies from the sample, where A'^g is the total 
number of galaxies in the sample. 

The successful, distinct predictions for the colors of 
central and satellite galaxies emerges naturally from the 
age distribution matching formalism. At fixed luminos- 
ity, the colors of our mock galaxies are drawn from the 
same color PDF, Psdss(<7 — ^\L), regardless of sub/host 
halo designation. Moreover, our color assignment algo- 
rithm takes no explicit account of (sub)halo mass. Thus 
the difference between the colors of satellites and cen- 
trals in our mock is purely a reflection of the different 
mass assembly history of host halos and subhalos. Ex- 
cepting only some mild tension for the color of satellite 
galaxies in a few of the brighter Mjr"^^ bins, the agreement 
between mock and observed central and satellite galaxy 
color is quite good. Again, we emphasize that we have 

^ We bootstrap resample with replacement, that is, we allow 
for the possibility of repeated draws of the same object. 

not tuned any parameters in our model and we have cho- 
sen to keep our implementation of the age distribution 
matching technique as simple as possible (see § 6 for an 
elaboration of this point). 


Mock catalogs based on age distribution matching inherit 
all the successes of abundance matching because this is 
precisely the method we use in the luminosity assignment 
phase of our algorithm. The color PDF Psdss(5 — r\L) 
plays the same role in the color assignment that the lumi- 
nosity function "I'sDSs(i) plays in the luminosity assign- 
ment. Thus just as $mock(i) from abundance matching 
agrees precisely with the observed luminosity function, 
so do our mock color distributions agree precisely with 
observed color PDF. 

There is an important conceptual difference between 
abundance matching and age distribution matching: the 
role that the property ^starve plays in the color assign- 
ment differs markedly from the role played by Vmax in 
the luminosity assignment. In abundance matching, lu- 
minosity assignment simply cannot proceed without ap- 
peal to the property Vmax (or some other halo property 
such as A'fvir for host halos and Mpcak for subhalos), be- 
cause this is an essential ingredient to the defining rela- 
tion ng(> L) — nh(> Vmax). This is not the case with 

© 0000 RAS, MNRAS 000, 000-000 

A. P Hearin & D. F Watson 

the color assignment. While the dark matter halo prop- 
erty ^starve is in monotonlc correspondence with [g — r) 
color just as Vmax is in monotonic correspondence with 
luminosity, this feature is not essential in order for the 
color assignment algorithm to proceed. Indeed, the rank- 
ordering oi [g — r) color with Zatarve only serves to corre- 
late galaxy color with halo mass assembly history, but a 
mock catalog constructed without this correlation would 
nonetheless have a color PDF that is in agreement with 
PsDSsig - r\L) . 

In light of this observation, one may wonder whether 
correlating galaxy color with ^starve is necessary at all in 
the construction of a successful model. For example, the 
color model in Skibba et al. (2007) has no such corre- 
lation. Instead, in order to correctly predict the 2PCF 
Skibba et al. (2007) draw [g — r) color for mock cen- 
trals and satellites from explicitly different distributions. 
While their model faithfully reproduces a variety of ob- 
served galaxy statistics, this class of models is unable 
even in principle to reproduce the observation that, at 
fixed luminosity, red central galaxies reside in more mas- 
sive host halos than blue centrals (More et al. 2011). This 
feature emerges naturally in the age distribution match- 
ing formalism, as we will show explicitly in a future com- 
panion paper. 

The expectation that ^starve should be related to stel- 
lar mass assembly is physically well-motivated. First, it 
is natural to expect that «form represents an important 
epoch in the history of the formation of a galaxy. As 
shown in Wechsler et al. (2002) , at Zform the dark matter 
halo transitions from the fast- to slow-accretion regime. 
After this epoch the matter accreting onto the dark mat- 
ter halo ceases to penetrate into the core of the halo where 
the galaxy resides^ , depriving the galaxy of new material 
from the field with which it can form stars. 

In reality, the halo may continue to form stars from 
its existing gas supply after this epoch, leading to a delay 
between ztor-m and the actual quenching of star formation. 
However, our model is based on rank- ordering of the prop- 
erty ^starve, and so the relative color of galaxies in halos 
with different formation times will be correct provided 
that this time delay is not a strong function of redshift 
or halo mass. As discussed in § 6.1, we will explore more 
complex models with quenching time delays in a future 
companion paper. We simply note here that corrections 
due to mass- and redshift-dependent time delays cannot 
be too severe or our model would not successfully predict 
the 2PCF and group-based statistics. 

Second, the existence of a characteristic halo mass 
Afchar above which star formation is highly inefficient also 
has strong empirical support. The halo mass-dependence 
of the stellar mass to halo mass ratio of galaxies shows 
a peak at Afchar ~ lQ^^h~^MQ that rapidly falls off as 
halo mass increases (Yang et al. 2012, 2013; Watson & 
Conroy 2013; Behroozi et al. 2012; Moster et al. 2013). 
This characteristic mass was recently shown to remain 
essentially constant for most of cosmic history (Behroozi 
et al. 2013). Additionally, Mchar ~ lO^^/i'^M© is the 
halo mass at which pressure-supported shocks can heat 

See their Figure 18 for a demonstration of this point. 

infalling gas to the virial temperature (Dekel & Birnboim 
2006). More importantly, this is also the mass when ef- 
fects due to AGN feedback become significant, which can 
have a dramatic effect on star formation (Teyssier et al. 
2011; Martizzi et al. 2012). 

In our implementation of age distribution matching, 
we frame no hypothesis for the particular physical mech- 
anism (s) that influence star formation inside massive ha- 
los. Instead, we simply posit that there exists a char- 
acteristic halo mass M^har above which star formation 
becomes inefficient, and that this leads to a correlation 
between galaxy color and the characteristic epoch Zchar 
that the halo attains this mass. 

We explored a range of masses for Mchar and find 
that both the 2PCF and our group statistics allow us to 
discriminate between competing values for the character- 
istic mass. In models where Mchar ^ lO^^'^/i^^M©, bright 
central galaxies are far too blue, and bright blue galaxies 
are strongly over-clustered on small scales (< lft~^Mpc). 
Conversely, in models with Mchar ^ IQ^^'^h^^ Mq central 
galaxies are too red and the 2PCF is in poor agreement 
with the data in all luminosity bins. 

Finally, we turn to the property Zacc, the epoch a 
satellite accreted onto its host halo. Numerous studies 
have shown that the dependence of galaxy color and/or 
star formation rate (SFR) has little-to-no residual en- 
vironmental dependence once small-scale environment 
(< lh~^yipc) has been accounted for (e.g., Hogg et al. 
2004; Blanton et al. 2005). This small-scale environmen- 
tal dependence has been shown to be driven primarily 
by correlations between the properties of the host halo 
and its satellite galaxies (Blanton & Berlind 2007; Wet- 
zel et al. 2011; Tinker et al. 2011; Wetzel et al. 2012). 
These results have given rise to a widely accepted picture 
in which satellite evolution is significantly influenced by 
processes that occur near or within the virial radius the 
host halo. 

There are a variety of physical mechanisms that 
occur inside the virial radius of a host halo that may 
have a significant impact on satellite galaxy color/SFR. 
These include, but are not limited to, mass loss due to 
outside-in stripping (see Watson et al. 2012, and ref- 
erences therein), "strangulation" (Larson et al. 1980; 
Diemand et al. 2007; Kawata & Mulchaey 2008), ram- 
pressure stripping (Gunn & Gott 1972; Abadi et al. 1999; 
McCarthy et al. 2008), "harassment" (Moore et al. 1998), 
and intra-host mergers (Makino & Hut 1997; Wetzel et al. 
2009a, b). Again, in our implementation of age distribu- 
tion matching, we do not preferentially select any one 
of these physical mechanisms as the primary cause for 
satellite quenching. Instead, we make the following sim- 
ple observation: the impact that each of the scenarios dis- 
cussed above has on satellite color/SFR will be greater 
for satellites that fell into their host halo potential at 
earlier times. Thus it is natural to expect a correlation 
between satellite galaxy color and Zacc. Accordingly, we 
identify Zacc as the third 'special' epoch in (sub)halo mass 
accretion history. 

Our treatment of the effect satellite accretion has 
on color/SFR is consistent with the finding in Wetzel 
et al. (2011) that there is no minimum host halo mass for 
post-accretion physics to significantly influence satellite 

© 0000 RAS, MNRAS 000, 000-000 

The Dark Side of Galaxy Color 9 

evolution. Moreover, since we correlate color with Zacc 
regardless of host halo mass, the success of our imple- 
mentation of age distribution matching provides new sup- 
porting evidence for conclusions drawn in van den Bosch 
et al. (2008), who found that satellite-specific processes 
are equally efficient in host halos of all masses. 

With the above physical picture in mind, in age 
distribution matching we make the following simple as- 
sumption: the earlier any one of the above processes he- 
gins in a halo, the redder its galaxy will be today. Math- 
ematically, this is formulated by introducing a mono- 
tonic correspondence between [g — r) color and ^starve = 


c} , as described in § 4.f .3. 

In a recent paper closely related to this work, Masaki 
et al. (20f3) paint a binary 'red' or 'blue' color onto 
(sub)halos that have first been assigned luminosities 
based on abundance matching. The authors explored a 
variety of different halo properties to use in the color as- 
signment. They found that several different choices result 
in successful predictions for the 2PCF, motivating their 
use of galaxy-galaxy lensing measurements to discrimi- 
nate between competing models. Since precision galaxy- 
galaxy lensing measurements that have been binned si- 
multaneously on both r-band luminosity and (g — r) color 
were not available, the authors instead used measure- 
ments binned on morphology and assumed a perfect cor- 
respondence between red (blue) galaxies and early (late) 
type morphological classification. 

While the assumption that all red (blue) galaxies 
have early (late) type morphology is a reasonable first 
approximation (Sheldon et al. 2004), we do not rely on 
it in this work. Instead, we employed galaxy group-based 
statistics to provide our additional discriminating power. 
Galaxy group membership yields information about halo 
occupation that is independent from 2PCF information. 
Group membership is necessarily identified in redshift 
space, and so group-based statistics incorporate informa- 
tion based on the velocity field, whereas projected clus- 
tering does not. Additionally, group statistics can probe 
halo occupation at higher masses than it is possible to do 
with small-scale clustering (see Hearin et al. 2012, Figure 
2). Nonetheless, in a future companion paper to this one, 
we will make new galaxy-galaxy lensing measurements of 
SDSS galaxies binned on {g — r) color to provide further 
tests of the success of the model we present here. 

The semi-analytic modeling (SAM) approach to pre- 
dicting galaxy properties such as color and star formation 
differs markedly from the theory presented here. SAMs 
also use the hierarchical build-up of dark matter halos 
as a foundation, but from here SAMs model the myr- 
iad baryonic processes that take place inside halos with 
a set of analytic formulae together with a large number 
of finely tuned parameters (White & Frenk 1991; Kauff- 
mann et al. 1999; Cole et al. 1994; Somerville & Primack 
1999; De Lucia & Blaizot 2007; Guo et al. 2011). Such ap- 
proaches have been quite successful at reproducing many 
statistics of the observed galaxy distribution, though the 
color-binned 2PCF has been notoriously difficult to cor- 
rectly predict in quantitative detail (see, e.g., figures 20 
& 21 of Guoet al. 2011). 

There have been several recent advances in the liter- 
ature on predicting galaxy color/SFR that dramatically 

simplify the set of assumptions and number of parame- 
ters required by most semi-analytic models. Mutch et al. 
(2013) couple dark matter halo accretion histories from 
the Millennium Simulation (Springel et al. 2005) with two 
simple analytic functions designed to encode the physics 
of baryonic growth and star formation history inside ha- 
los (see also Cattaneo et al. 2011, for a closely related 
formulation). An alternative approach is taken in Lilly 
et al. (2013), in which the authors predicate their the- 
ory upon the assumption that the gas mass in a galaxy's 
reservoir instantaneously regulates star formation. 

In an alternative approach that shares many concep- 
tual similarities to our own, the model studied in Peng 
et al. (2010) begins with a minimal set of phenomenolog- 
ical observations based on zCOSMOS data (Lilly et al. 
2007) and apply a set of continuity equations to the evo- 
lution of the number of red and blue galaxies in order to 
connect galaxy properties to the dark matter halos host- 
ing them. The "separability condition" discussed in Peng 
et al. (2010) is reflected by our formulation of the prop- 
erty Zstarvc, iu that Zchar , Zform, and Zacc are calculated 
purely independently for each halo. 

Following Peng et al. (2012) we define "mass quench- 
ing" as that set of processes that are not satellite-specific, 
so that in our theory (sub)halos with Zstarvc > ^acc are 
said to be mass-quenched, and subhalos with ^starve = 
Zacc are "environment-quenched" . The success of age dis- 
tribution matching has revealed a new channel of mass 
quenching. The manner in which we implement ^char 
plays a virtually identical role to the model of mass 
quenching explored in Peng et al. (2012), who suppose 
that star formation rapidly quenches after a halo achieves 
a certain mass Mquench- The necessity of the role played 
by 2form in our model strongly suggests that the rate of 
mass accretion in a halo plays a role of comparable im- 
portance to Mquench iu uiass quenching. 

Each of the above studies (that is. Mutch et al. 2013; 
Lilly et al. 2013; Peng et al. 2012) assumes a small set 
of simple analytic formulae together with knowledge of 
structure formation in ACDM to accurately reproduce 
a wide variety of one-point statistics of the galaxy dis- 
tribution. By contrast, age distribution matching uses 
one-point statistics as input'^ and correctly reproduces 
the two-point function (Fig. 2) as well as higher-order 
statistics based on galaxy groups (Fig. 3). The key to the 
success of our theory is the identification of a special time 
^starve m each halo's assembly history that is physically 
associated with the slowing and ultimate quenching of star 
formation in the galaxy at the center of the halo. 

6.1 Future Work 

The primary aim of this paper is to introduce a new the- 
oretical framework for studying the galaxy-halo connec- 
tion, and to show that this model is successful even with 
the relatively simple implementation we have presented 
here. Accordingly, wherever possible we have opted to 
employ simplifying approximations that are widely used 

^ Specifically the luminosity function "l'sDSs(i) ^^'^ the dis- 
tribution of colors PsDSs(9 ~ ''|i)- 

© 0000 RAS, MNRAS 000, 000-000 

10 A. P Hearin & D. F Watson 

in the literature rather than tuning our exploration of 
halo mass accretion history to provide better fits to the 
observations. For example, as discussed in Appendix A 
we assume that the formation history of dark matter ha- 
los can be encompassed by a single parameter, Zform, and 
we appeal to the results in Wechsler et al. (2002) in our 
use of halo concentration as a proxy for Zform- In this 
section we call direct attention to some of these simpli- 
fying assumptions, and discuss possible extensions of our 
model that we intend to explore in future work. 

As discussed at length in Wetzel et al. (2011), the ro- 
bust bimodality of the specific star formation rate (SSFR) 
distribution at low-redshift constitutes compelling evi- 
dence that post-accretion quenching of satellite star for- 
mation occurs rapidly after substantial time delay (~ 
2 — 4Gyr). This motivates a modification to the formalism 
we introduce here, in which rather than, the {g — r) 
color of environment-quenched satellites is instead corre- 
lated with 2;quonch = 2acc + Atdoiay ■ We have conducted 
a preliminary investigation of such models, finding that 
Atdeiay = 2Gyr has little effect on either the 2PCF or the 
group statistics. However, the physical processes leading 
to a correlation between color and Zchar may also have 
a delayed onset, for example due to the length of time 
required for a pressure-supported shock to propagate to 
the virial radius. Since the physics underlying these two 
delays {At'^ily and At^li^^) is Ukely quite distinct, there 
is no reason to assume that they are equal. We postpone 
a more systematic investigation of this parameter space 
for the companion paper mentioned above. 

In abundance matching, there is scatter between lu- 
minosity and Vmax. In this work, we have assumed a 
purely monotonic relation between {g — r) color and the 
property ^starve, but it is easy to imagine implementations 
of age distribution matching in which there is scatter in 
this relation as well. A simple way to accomplish this 
would be an adaptation of the scatter method detailed 
in Appendix A of Hearin et al. (2012). Briefly, one would 
simply introduce Gaussian noise in the randomly drawn 
{g — r) colors and use the rankings of the noisy [g — r) 
values to appropriately shuffle the Zstarvc— ranking of the 
(sub)halos. Again, we have not pursued this complica- 
tion because our present aim is to show that even the 
simplest formulation of age distribution matching is re- 
markably successful. However, we intend to explore such 
extensions in future investigations of this technique, par- 
ticularly when color-binned galaxy-galaxy lensing mea- 
surements become available. 

In this paper, we have only investigated our theory 
at low redshift. However, if the luminosity function and 
color PDF are well measured at other epochs then our 
model should be consistent across redshifts, and this a 
crucial next step for testing the generality of our age dis- 
tribution matching formalism. Furthermore, our formal- 
ism generalizes in the obvious way to mock galaxies with 
stellar mass and/or (S)SFR instead of luminosity. We in- 
tend to study mocks based on such properties in the fu- 
ture as these are fundamental galaxy properties and are 
more directly linked to color. Such tests will be enabled 
by existing 2PCF meeisurements as a function of stellar 
mass split on blue and red galaxy populations at z ~ 

(Li et al. 2006; Jiang et al. 2011) and also z ~ 1 (Mostek 
et al. 2012). 

Our model is also general in the sense that the first 
phase of the algorithm need not necessarily be abundance 
matching. For example, one could instead begin by using 
the CLF formalism to assign brightnesses to dark mat- 
ter halos. Alternatively, a stellar mass to halo mass map 
could be applied to construct a mock based on stellar 
masses. In either case, the color assignment phase of the 
algorithm would proceed in an identical fashion to what 
is described in S 4.1.3. 


In this section we summarize our primary results and 

1. We introduce age distribution matching, a new the- 
oretical framework for studying the relationship between 
halo mass accretion history and stellar mass assembly. 

2. By construction, over the entire luminosity range of 
our sample the {g — r) color PDF of mock galaxies based 
on age distribution matching is in exact agreement with 

3. Our model successfully predicts luminosity- and 
color-binned two-point projected galaxy clustering at 
low-redshift, as well as central and satellite galaxy color 
as a function of host halo environment [M^'^^). We em- 
phasize that these are predictions of the theory, and that 
we have not relied upon any fitting or fine-tuning of any 
model parameters. The success of our model implies that 
galaxy and halo assembly are strongly correlated, in con- 
trast to recent claims in the literature. 

4. We make publicly available 

• Our A/r < — 19 mock galaxy catalog. 

• Full mass accretion histories of the main progeni- 
tors of all 2 = Bolshoi halos. 


APH and DFW are particularly grateful to Audrey 
Kravtsov for invaluable feedback throughout the develop- 
ment of this work. We also thank Matt Becker, Benedikt 
Diemer, Simon Lilly, Jeff Newman, Andrew Zentner, and 
Nick Gnedin for useful discussions at various stages, and 
Charlie Conroy and Frank van den Bosch for comments 
on an early draft. We thank Andreas Berlind for the use 
of his group finder and for early discussions about this 
type of modeling. We thank Peter Behroozi for making 
his halo catalogs and merger trees publicly available, and 
also for reliably prompt and clear answers to questions 
about these catalogs. We thank Ramin Skibba for shar- 
ing his mock catalog with us and for many informative 
discussions. Finally, we thank John Fahey for Transfigu- 
rations of Blind Joe Death. 

APH is also supported by the U.S. Department of 
Energy under contract No. DE-AC02-07CH11359. DFW 
is supported by the National Science Foundation under 
Award No. AST-1202698. 

© 0000 RAS, MNRAS 000, 000-000 

The Dark Side of Galaxy Color 11 


Abadi M. G., Moore B., Bower R. G., 1999, MNRAS , 

308, 947 
Abazajian K. N., Adelman-McGarthy J. K., Agiieros 

M. A., AUam S. S., AUende Prieto G., An D., Anderson 

K. S. J., Anderson S. F., Annis J., Bahcall N. A., et al. 

2009, ApJS , 182, 543 
Baldry I. K., Glazebrook K., Brinkmann J., Ivezic Z., 

Lupton R. H., Nichol R. C., Szalay A. S., 2004, ApJ , 

600, 681 
Balogh M. L., Morris S. L., Yee H. K. G., Garlberg R. G., 

EUingson E., 1999, ApJ , 527, 54 
Behroozi R S., Gonroy G., Wechsler R. H., 2010, ApJ , 

717, 379 
Behroozi P. S., et al., 2013a, ApJ , 763, 18 
Behroozi P. S., et al., 2013b, ApJ , 762, 109 
Behroozi P. S., Wechsler R. H., Gonroy G., 2012, 

Behroozi P. S., Wechsler R. H., Gonroy G., 2013, ApJL , 

762, L31 
Bell E. F., et al., 2004, ApJ , 608, 752 
Berlind A. A., et al., 2006, ApJS , 167, 1 
Berlind A. A., Weinberg D. H., 2002, ApJ , 575, 587 
Blanton M. R., Berlind A. A., 2007, ApJ , 664, 791 
Blanton M. R., Eisenstein D., Hogg D. W., Schlegel 

D. J., Brinkmann J., 2005, ApJ , 629, 143 
Blanton M. R., et al., 2003, ApJ , 594, 186 
Blumenthal G. R., Faber S. M., Primack J. R., Rees 

M. J., 1984, Nature , 311, 517 
Bryan G. L., Norman M. L., 1998, ApJ , 495, 80 
Gacciato M., van den Bosch F. G., More S., Mo H., Yang 

X., 2013, MNRAS , 430, 767 
GaroUo C. M., et al., 2012, ArXiv: 1206.5807 
Gattaneo A., Mamon G. A., Warnick K., Knebe A., 

2011, AAP, 533, A5 
Gole S., Aragon-Salamanca A., Frank C. S., Navarro 

J. F., Zepf S. E., 1994, MNRAS , 271, 781 
Gonroy G., Wechsler R. H., 2009, ApJ , 696, 620 
Gonroy G., Wechsler R. H., Kravtsov A. V., 2006, ApJ , 

647, 201 
Gooper M. C., et al., 2006, MNRAS , 370, 198 
Gooper M. C., et al., 2012, MNRAS , 419, 3018 
Gooray A., Sheth R., 2002, Phys. Rep. , 372, 1 
De Lucia G., Blaizot J., 2007, MNRAS , 375, 2 
Dekel A., Birnboim Y., 2006, MNRAS , 368, 2 
Diemand J., Kuhlen M., Madau P., 2007, ApJ , 667, 859 
Gerke B. F., Wechsler R. H., Behroozi P. S., Cooper 

M. G., Yan R., Goil A. L., 2012, ArXiv: 1207.2214 
Gottloeber S., Klypin A., 2008, ArXiv:0803.4343 
Gunn J. E., Gott III J. R., 1972, ApJ , 176, 1 
Guo H., Zehavi I., Zheng Z., 2012, ApJ , 756, 127 
Guo Q., White S., Boylan-Kolchin M., De Lucia G., 

KaufFmann G., Lemson G., Li G., Springel V., Wein- 

mann S., 2011, MNRAS , 413, 101 
Guo Q., White S., Li G., Boylan-Kolchin M., 2010, MN- 
RAS , 404, 1111 
Hearin A. P., Zentner A. R., Berlind A. A., Newman 

J. A., 2012, ArXiv:1210.4927 
Hearin A. P., Zentner A. R., Newman J. A., Berlind 

A. A., 2012, ArXiv: 1207. 1074 
Hogg D. W., et al., 2004, ApJL , 601, L29 

Jiang T., Hogg D. W., Blanton M. R., 2011, 

ArXiv: 1104.5483 
Kauffmann G., Golberg J. M., Diaferio A., White 

S. D. M., 1999, MNRAS , 303, 188 
Kawata D., Mulchaey J. S., 2008, ApJL , 672, L103 
Klypin A. A., Trujillo-Gomez S., Primack J., 2011, ApJ , 

740, 102 
Krause E., Hirata G. M., Martin C., Neill J. D., Wyder 

T. K., 2013, MNRAS , 428, 2548 
Kravtsov A. V., 2013, ApJL , 764, L31 
Kravtsov A. V., Berlind A. A., Wechsler R. H., Klypin 

A. A., Gottlober S., AUgood B., Primack J. R., 2004, 

ApJ , 609, 35 
Kravtsov A. V., Klypin A. A., Khokhlov A. M., 1997, 

ApJS , HI, 73 
Larson R. B., Tinsley B. M., Galdwell G. N., 1980, ApJ , 

237, 692 
Leauthaud A., et al., 2011, ArXiv: 1104.0928 
Li C., Kauflmann G., Jing Y. P., White S. D. M., Borner 

G., Cheng F. Z., 2006, MNRAS , 368, 21 
Li C, White S. D. M., 2009, ArXiv:0910.2982 
Lilly S. J., GaroUo CM., Pipino A., Renzini A., Peng 

Y., 2013, ArXiv: 1303.5059 
Lilly S. J., et al., 2007, ApJS , 172, 70 
Makino J., Hut P., 1997, ApJ , 481, 83 
Martizzi D., Teyssier R., Moore B., 2012, MNRAS , 420, 

Masaki S., Lin Y.-T., Yoshida N., 2013, ArXiv e-prints 
McCarthy I. G., Frenk C. S., Font A. S., Lacey G. G., 

Bower R. G., Mitchell N. L., Balogh M. L., Theuns T., 

2008, MNRAS , 383, 593 
Moore B., Lake G., Katz N., 1998, ApJ , 495, 139 
More S., van den Bosch F. G., Gacciato M., More A., 

Mo H., Yang X., 2013, MNRAS , 430, 747 
More S., van den Bosch F. G., Gacciato M., Skibba R., 

Mo H. J., Yang X., 2011, MNRAS, 410, 210 
Mostek N., Goil A. L., Gooper M. C, Davis M., Newman 

J. A., Weiner B., 2012, ArXiv:1210.6694 
Moster B. P., Naab T., White S. D. M., 2013, MNRAS , 

428, 3121 
Moster B. P., Somerville R. S., Maulbetsch C, van den 

Bosch F. G., Maccio A. V., Naab T., Oser L., 2010, 

ApJ , 710, 903 
Mutch S. J., Croton D. J., Poole G. B., 2013, 

ArXiv: 1304.2774 
Neistein E., Weinmann S. M., Li G., Boylan-Kolchin M., 

2010, ArXiv:1011.2492 
Norberg P., et al., 2002, MNRAS , 332, 827 
Peacock J. A., Smith R. E., 2000, MNRAS , 318, 1144 
Peng Y.-j., et al., 2010, ApJ , 721, 193 
Peng Y.-j., Lilly S. J., Renzini A., GaroUo M., 2012, 

ApJ , 757, 4 
Pizagno J., et al., 2007, A J , 134, 945 
Reddick R. M., Wechsler R. H., Tinker J. L., Behroozi 

P. S., 2012, ArXiv:1207.2160 
Riebe K., Parti A. M., Enke H., Forero-Romero J., 

Gottloeber S., Klypin A., Lemson G., Prada F., Pri- 
mack J. R., Steinmetz M., Turchaninov V., 2011, 

ArXiv: 1109.0003 
Rodriguez-Puebla A., Drory N., Avila-Reese V., 2012, 

ApJ , 756, 2 

© 0000 RAS, MNRAS 000, 000-000 

12 A. P Hearin & D. F Watson 

Scoccimarro R., Sheth R. K., Hui L., Jain B., 2001, ApJ , 

546, 20 
Seljak U., 2000, MNRAS , 318, 203 
Sheldon E. S., Johnston D. E., Frieman J. A., Scranton 

R., McKay T. A., Connolly A. J., Budavari T., Zehavi 

I., Bahcall N. A., Brinkmann J., Fukugita M., 2004, 

AJ , 127, 2544 
Simha V., Weinberg D., Dave R., Fardal M., Katz N., 

Oppenheimer B. D., 2010, ArXiv:1011.4964 
Skibba R. A., Sheth R. K., 2009, MNRAS , 392, 1080 
Skibba R. A., Sheth R. K., Martino M. C, 2007, MN- 
RAS , 382, 1940 
Somerville R. S., Primack J. R., 1999, MNRAS , 310, 

Springel V., White S. D. M., Jenkins A., Frenk C. S., 

Yoshida N., Gao L., Navarro J., Thacker R., Croton D., 

Helly J., Peacock J. A., Cole S., Thomas P., Couchman 

H., Evrard A., Colberg J., Pearce F., 2005, Nature , 

435, 629 
Tasitsiomi A., Kravtsov A. V., Wechsler R. H., Primack 

J. R., 2004, ApJ , 614, 533 
Teyssier R., Moore B., Martizzi D., Dubois Y., Mayer 

L., 2011, MNRAS , 414, 195 
Tinker J., Wetzel A., Conroy C, 2011, ArXiv:1107.5046 
Tinker J. L., Wetzel A. R., 2010, ApJ , 719, 88 
Vale A., Ostriker J. P., 2004, MNRAS , 353, 189 
Vale A., Ostriker J. P., 2006, MNRAS , 371, 1173 
van den Bosch F. C, Aquino D., Yang X., Mo H. J., 

Pasquali A., Mcintosh D. H., Weinmann S. M., Kang 

X., 2008, MNRAS , 387, 79 
van den Bosch F. C, Yang X., Mo H. J., Weinmann 

S. M., Maccio A. V., More S., Cacciato M., Skibba R., 

Kang X., 2007, MNRAS , 376, 841 
Watson D. F., Berlind A. A., Zentner A. R., 2012, ApJ , 

754, 90 
Watson D. F., Conroy C, 2013, ArXiv: 1301. 4497 
Wechsler R. H., Bullock J. S., Primack J. R., Kravtsov 

A. v., Dekel A., 2002, ApJ , 568, 52 
Weinmann S. M., Kauffmann G., van den Bosch F. C, 

Pasquah A., Mcintosh D. H., Mo H., Yang X., Guo Y., 

2009, MNRAS , 394, 1213 
Weinmann S. M., van den Bosch F. C, Yang X., Mo 

H. J., 2006, MNRAS , 366, 2 
Wetzel A. R., Cohn J. D., White M., 2009a, MNRAS , 

395, 1376 
Wetzel A. R., Cohn J. D., White M., 2009b, MNRAS , 

394, 2182 
Wetzel A. R., Tinker J. L., Conroy C, 2011, 

Wetzel A. R., Tinker J. L., Conroy C, van den Bosch 

F. C, 2012, ArXiv:1206.3571 
Wetzel A. R., Tinker J. L., Conroy C, van den Bosch 

F. C, 2013, ArXiv:1303.7231 
Wetzel A. R., White M., 2010, MNRAS , 403, 1072 
Whitaker K. E., et al., 2011, ApJ , 735, 86 
White S. D. M., Frenk C. S., 1991, ApJ , 379, 52 
White S. D. M., Rees M. J., 1978, MNRAS , 183, 341 
Yang X., Mo H. J., van den Bosch F. C, 2003, MNRAS , 

339, 1057 
Yang X., Mo H. J., van den Bosch F. C, Bonaca A., Li 

S., Lu Y., Lu Y., Lu Z., 2013, ArXiv:1302.1265 

Yang X., Mo H. J., van den Bosch F. C, Zhang Y., Han 

J., 2011, ArXiv:1110.1420 
Yang X., Mo H. J., van den Bosch F. C, Zhang Y., Han 

J., 2012, ApJ , 752, 41 
York D. C, et al., 2000, AJ , 120, 1579 
Zehavi I., et al., 2002, ApJ , 571, 172 
Zehavi I., et al., 2005, ApJ , 630, 1 
Zehavi I., et al., 2011, ApJ , 736, 59 
Zheng Z., Coil A. L., Zehavi I., 2007, ApJ , 667, 760 


The central tenet of age distribution matching is that 
galaxy color is determined by halo mass accretion history, 
at least in a statistical sense. In this appendix we discuss 
the details of how we calculate Zacc , Zform , and Zchar , the 
key quantities used in the color assignment phase of our 

Zacc : In accordance with the convention adopted 
in Behroozi et al. (2013a), we use the snapshot after 
which the subhalo always remains a subhalo. We also 
explored the definition of accretion advocated in Wetzel 
et al. (2013), where subhalos are said to have accreted 
the first time they have been identified as subhalos for 
two consecutive time-steps, and find that this choice has 
little impact on our results. 

Zform : We use the approximation introduced in 
Wechsler et al. (2002), who find that mass accretion his- 
tory is well-fit by an exponential, Mvir(^) = Moe~°"^ . The 
authors in Wechsler et al. (2002) defined the scale factor 
of halo formation aform to be the epoch when the logarith- 
mic slope 5* of Mvir(2) drops below S = 2. While there 
is some amount of arbitrariness in this particular chosen 
value of S, the transition that occurs at or near this time 
has very real physical consequences: the central density 
of the halo ceases to increase (see their Figure 18). This 
physical change naturally gives rise to a correlation be- 
tween halo formation time and NFW concentration Cvir 
(defined as the ratio of Rviv/vs). 

One of the central results of Wechsler et al. (2002) is 
that ttform is well-approximated by the following equation: 


': 4.1ao/Cvir, 

where Zform = l/flform — 1, aud ao is the scale factor at 
the time of "observation" . For host halos we take the time 
of observation to be ao = 1 and use the halo's concen- 
tration Cvir at this snapshot. Tidal disruption and mass 
stripping of subhalos after accretion makes ao = 1 esti- 
mates of i?vir noisy and unreliable. Fortunately, Wechsler 
et al. (2002) showed that the scaling of halo formation 
with Cvir remains the same over a broad range of ao. So 
for subhalos, we define Cvir to be the NFW concentration 
of the subhalo at ao = aacc, the time of accretion (as de- 
fined above). We note that for ~ 10% of subhalos, this 
definition produces an unphysically long tail at high red- 
shift in the aform distribution. We attribute this to large, 
transient boosts to the subhalo concentration at aacc due 
to a tidal event. However, we find that our results are 
insensitive to this behavior: when we collapse all objects 
in this tail to Zform = 3 we get virtually identical results. 

© 0000 RAS, MNRAS 000, 000-000 

The Dark Side of Galaxy Color 13 



We use the first time a (sub)lialo attains 

0. As noted in tiie main body of tlie paper, we 

and found 10^^ ft" ^Mq 

explored a range of values of Afch, 
is the most successful. 

© 0000 RAS, MNRAS 000, 000-000